Regional intrusion identification method and system under view angle of unmanned aerial vehicle

Abstract

The invention relates to the technical field of regional security and protection monitoring, in particular to a regional intrusion identification method and system under the view angle of an unmanned aerial vehicle, which adopts a multi-source information fusion framework, combines the real-time position, attitude and flight state information of the unmanned aerial vehicle through an accurate space coordinate conversion algorithm, and realizes the regional intrusion identification under the view angle of the unmanned aerial vehicle. GIS geographic information, unmanned aerial vehicle monitoring and a deep learning detection result are deeply fused, high-precision positioning of an intrusion area in an image shot by an unmanned aerial vehicle is realized, and robustness in a complex environment is improved. Targets such as personnel and vehicles in the image are efficiently recognized by applying a YOLO series algorithm, and violation intrusion behaviors are accurately judged by calculating the intersection-to-union ratio IoU of the targets and the intrusion area. The unmanned aerial vehicle is utilized to quickly cover a large-area monitoring area, and the problem that the monitoring range of a traditional fixed camera is limited is solved. In addition, by means of a space coordinate conversion algorithm, the influence caused by height and angle changes in the flight of the unmanned aerial vehicle is effectively overcome, and accurate positioning of the intrusion area in the image is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of regional security monitoring, in particular to a regional intrusion identification method and system under the perspective of a UAV. BACKGROUND

[0002] With the rapid development of artificial intelligence technology and the increasing demand for security protection, security monitoring technology is also facing new challenges and expectations. In traditional regional intrusion identification technology, the application of fixed cameras dominates, but the monitoring range of fixed cameras is relatively small and the position and shooting angle of the camera are usually difficult to adjust flexibly after installation. The introduction of UAV technology provides a new idea to solve the limitations of fixed cameras. UAVs can fly flexibly in the air and take pictures from different angles, thus obtaining comprehensive and rich image information.

[0003] However, in actual application, the high-speed flight characteristics of the UAV result in rapid changes in the shooting picture, which makes it difficult for traditional regional calibration methods based on fixed cameras to be directly applicable to UAV scenes. When identifying a specific region in the picture, the existing target detection technology is prone to produce a large number of misidentification results if the area of the intrusion region to be identified is small or its features are not prominent, and the positioning accuracy is difficult to meet the actual demand, which cannot accurately determine whether the target has truly entered the specific intrusion region, thereby seriously affecting the accuracy and reliability of regional intrusion identification. Moreover, the height, angle and shooting angle of the UAV inevitably change during flight, and these factors together make it extremely difficult to accurately determine the position of the intrusion region in the image. At the same time, most of the existing regional intrusion identification technologies fail to fully integrate the advantages of GIS and UAV monitoring. GIS systems contain rich geographic spatial information such as topography, land attribute, etc., but fail to deeply integrate with image data obtained by UAVs, resulting in a lack of accuracy and consistency in the calibration and identification of intrusion regions in geographic space, and making it difficult to achieve high-precision and high-reliability regional intrusion identification. SUMMARY

[0004] The technical problem solved by the present application is to provide a regional intrusion identification method and system under the perspective of a UAV to solve the technical problem of insufficient accuracy caused by unobvious regional features and changes in the pose of the UAV in the existing UAV regional calibration process.

[0005] The basic scheme provided by the present application is a regional intrusion identification method under the perspective of a UAV, comprising the steps of:

[0006] S1: calibrating a polygonal intrusion region in a GIS map to obtain geographic coordinate information of the region vertices;

[0007] S2: The UAV collects video images and its pose parameters in real time;

[0008] S3: Based on the geographic coordinate information and the pose parameters, the pixel coordinates of the intrusion area in the UAV image are calculated through a spatial coordinate conversion algorithm;

[0009] S4: The target class is identified from the video images in real time through a target detection model, and the detection box pixel coordinate information of the target is output;

[0010] S5: The intersection over union (IoU) of the target detection box and the intrusion area pixel coordinates is calculated, and if the IoU exceeds a preset threshold, it is determined as a violation intrusion;

[0011] S6: When there is a violation intrusion behavior, an alarm is sent to the relevant personnel, and evidence data is generated according to the video images of the violation intrusion behavior.

[0012] Further, the S1 comprises:

[0013] S11: A closed polygon intrusion area is drawn in the GIS map through interactive operation;

[0014] S12: The geographic coordinate information of each vertex of the region includes longitude, latitude and altitude.

[0015] Further, the spatial coordinate conversion algorithm in S3 comprises:

[0016] S31: Convert the geographic coordinates to a 3D Cartesian coordinate system;

[0017] S32: Combine the UAV pose parameters and the camera field of view angle of the UAV-mounted camera to calculate the pixel mapping value of the intrusion area vertex in the image;

[0018] S33: Filter the effective mapping coordinates through a preset determination condition.

[0019] Further, the calculation of the pixel mapping value comprises:

[0020] S321: According to the straight-line distance between the UAV and the region vertex, the height difference and the camera parameters, the included angle between the target connecting line and the horizontal / vertical field of view angle is calculated;

[0021] S322: Based on the included angle and the image resolution, the pixel coordinates are generated.

[0022] Further, the S5 comprises:

[0023] S51: Calculate the overlapping area ratio IoU of the target detection box and the intrusion area:

[0024]

[0025] S52: When the IoU exceeds the preset threshold, an unauthorized intrusion is determined.

[0026] Furthermore, the target detection model is a YOLO series algorithm, and the objects to be identified include, but are not limited to, people and vehicles.

[0027] Furthermore, the alarm methods of S6 include audible and visual alarms, SMS notifications, and network message pushes.

[0028] A regional intrusion identification system from the perspective of an unmanned aerial vehicle (UAV), characterized in that it includes:

[0029] GIS area labeling module: Provides a map interactive interface for labeling intrusion areas;

[0030] UAV data acquisition module: acquires video streams and pose parameters in real time;

[0031] Coordinate mapping module: Performs the conversion from geographic coordinates to image pixels;

[0032] Target recognition module: Detects targets based on the YOLO algorithm;

[0033] Intrusion detection module: Determines intrusion behavior through IoU calculation;

[0034] Alarm module: Triggers alarms in real time and stores evidence.

[0035] Furthermore, the area intrusion identification method from any of the aforementioned drone perspectives was applied.

[0036] The principles and advantages of this invention are as follows: This invention employs a multi-source information fusion framework. Through a precise spatial coordinate transformation algorithm, combined with the real-time position, attitude, and flight status information of the UAV, it deeply integrates GIS geographic information, UAV monitoring, and deep learning detection results to achieve high-precision positioning of the intrusion area in UAV-captured images, improving robustness in complex environments. It utilizes YOLO series algorithms to efficiently identify targets such as people and vehicles in images, and accurately determines unauthorized intrusion behavior by calculating the Intersection over Union (IoU) between the target and the intrusion area.

[0037] Compared with existing technologies, the area intrusion identification method from the perspective of drones in this invention utilizes the flexibility and high-altitude perspective of drones to quickly cover a large area of ​​monitoring, effectively solving the problem of limited monitoring range of traditional fixed cameras. It is especially suitable for complex terrain, open areas and hard-to-reach areas.

[0038] In terms of positioning accuracy, by integrating GIS map information with real-time UAV status data and using spatial coordinate transformation algorithms, the influence of changes in altitude and angle during UAV flight is effectively overcome, achieving accurate positioning of the intrusion area in the image and avoiding misjudgment due to positioning deviation. Compared with traditional methods, this significantly improves the accuracy of regional intrusion identification.

[0039] From the perspective of application cost and flexibility, the application of drones reduces the equipment and maintenance costs associated with the large-scale deployment of traditional fixed cameras. Furthermore, this method allows for flexible adjustment of intrusion areas on the GIS map according to actual needs, adapting to the monitoring requirements of different scenarios. Real-time alarm and evidence retention functions enable timely responses to unauthorized intrusions, providing strong data support for security incident handling, significantly improving the efficiency and intelligence of security monitoring, and meeting the growing security demands of modern society. Attached Figure Description

[0040] Figure 1 This is a flowchart illustrating the steps of a method for identifying regional intrusions from the perspective of an unmanned aerial vehicle (UAV) according to the present invention.

[0041] Figure 2 This is a flowchart of spatial coordinate transformation for an embodiment of a method for regional intrusion identification from the perspective of an unmanned aerial vehicle (UAV) according to the present invention.

[0042] Figure 3 This is a logic block diagram of a second embodiment of a method for identifying regional intrusions from the perspective of an unmanned aerial vehicle (UAV) according to the present invention. Detailed Implementation

[0043] The following detailed description illustrates the specific implementation method:

[0044] The specific implementation process is as follows:

[0045] Example 1

[0046] Example 1 is attached. Figure 1 As shown, a method for regional intrusion identification from the perspective of an unmanned aerial vehicle (UAV) includes the following steps:

[0047] S1: Mark the polygonal invasion area on the GIS map and obtain the geographic coordinate information of the area's vertices;

[0048] S2: The drone collects video images and its own pose parameters in real time;

[0049] S3: Based on the geographic coordinate information and pose parameters, calculate the pixel coordinates of the intrusion area in the UAV image using a spatial coordinate transformation algorithm;

[0050] S4: Identify target categories from video images in real time using a target detection model, and output the pixel coordinates of the target's detection box;

[0051] S5: Calculate the intersection-union ratio (IoU) between the target detection box and the pixel coordinates of the intrusion area. If the IoU exceeds the preset threshold, it is determined to be an unauthorized intrusion.

[0052] S6: When unauthorized intrusion occurs, an alarm is issued to relevant personnel and evidence data is generated based on the video images used to determine the unauthorized intrusion.

[0053] By employing a precise spatial coordinate transformation algorithm, combined with real-time UAV position, attitude, and flight status information, and deeply integrating GIS geographic information, UAV monitoring, and deep learning detection results, high-precision localization of intrusion areas in UAV-captured images is achieved, enhancing robustness in complex environments. A target detection model efficiently identifies targets such as people and vehicles in the images, and by calculating the Intersection over Union (IoU) between the target and the intrusion area, unauthorized intrusion behavior is accurately determined.

[0054] Specifically, this solution first identifies the intrusion area using GIS mapping. Closed polygons (such as those for reservoir restricted areas) are drawn through the GIS interactive interface, and the latitude and longitude of the vertices (E108°30′12″, N34°15′30″) and altitude (520m) are obtained. The system automatically verifies the closure of the polygons, discarding invalid coordinates (such as longitude > 180°). The geographic coordinate information of each vertex in the area, including latitude, longitude, and altitude, is stored in the database, supporting dynamic updates. This solves the problem of inaccurate area identification in traditional methods; closure verification prevents algorithm crashes, such as mapping failures caused by unclosed polygons in existing technologies; and dynamic updates adapt to changes in security requirements.

[0055] Data is then collected via drones. In this embodiment, a DJI M300 drone equipped with an H20T camera (3840×2160 resolution) cruises along a preset path. An RTK positioning module is used to acquire latitude and longitude (±1cm accuracy), pitch angle (-90°~+30°), and yaw angle (0°~360°) in real time. The collected video images and pose parameters are transmitted back to the control center via a 5G network. Using high-precision pose parameters as the basis for coordinate mapping provides higher positioning accuracy compared to traditional GPS positioning (±5m error).

[0056] Then, based on the aforementioned geographic coordinate information and pose parameters, the pixel coordinates of the intrusion area in the UAV image are calculated using a spatial coordinate transformation algorithm. This spatial coordinate transformation algorithm includes: converting the geographic coordinates to a 3D Cartesian coordinate system; combining the UAV pose parameters and the field of view of the UAV's camera; and calculating the pixel mapping values ​​of the intrusion area vertices in the image. Valid mapping coordinates are then selected using preset criteria. Details are attached. Figure 2 As shown, this embodiment includes the following steps:

[0057] ①. Convert latitude and longitude to radians using the radian conversion formula:

[0058] lon1 rad =lon1×(π / 180)

[0059] lat1 rad =lat1×(π / 180)

[0060] lon2 rad =lon2×(π / 180)

[0061] lat2 rad =lat²×(π / 180)

[0062] In the formula, lon1 represents the latitude and longitude of the UAV, and lon2 and lat2 represent the latitude and longitude of the region vertices;

[0063] ②. Calculate the angle between the line connecting the UAV and the region vertex and due north. First, calculate the difference in radians in longitude. The calculation formula is as follows:

[0064] delta lon =lat2 rad -lon1 rad

[0065] Then, the angle between the line connecting the UAV and the vertex of the region and due north is calculated using the azimuth formula:

[0066] yy = sin(delta) lon )×cos(lon2 rad )

[0067] xx=cos(lat1 rad )×sin(lat2 rad )-sin(lat1 rad )×cos(lat2 rad )×cos(delta lon )

[0068] angle=arctan(yy / xx)×(180 / π)

[0069] Here, xx and yy represent the components of the line connecting the UAV and the region vertex on the x and y axes, respectively, and angle represents the angle between it and the due north direction.

[0070] ③. Calculate the straight-line distance between the drone and the vertex of the region. First, convert the latitude and longitude to a 3D Cartesian coordinate system. The formula is:

[0071] x1=(R+h1)×cos(lat1 rad)×cos(log1 rad )

[0072] y1=(R+h1)×cos(lat1 rad )×sin(lon1 rad )

[0073] z1=(R+h1)×sin(lat1 rad )

[0074] x2=(R+h2)×cos(lat2 rad )×cos(lon2 rad )

[0075] y2=(R+h2)×cos(lat2 rad )×sin(lon2 rad )

[0076] z2=(R+h2)×sin(lat2 rad )

[0077] In the formula, lon1 rad and lat1 rad This represents the radian value corresponding to the latitude and longitude of the drone, lon2 rad and lat2 rad The values ​​represent the radian values ​​corresponding to the latitude and longitude of the region vertex, R represents the Earth's radius, h1 and h2 represent the altitudes of the drone and the region vertex, respectively, x1, y1, and z1 represent the coordinates of the drone converted to 3D Cartesian coordinates, and x2, y2, and z2 represent the coordinates of the region vertex converted to 3D Cartesian coordinates.

[0078] Then, the distance between the drone and the region vertex is obtained using the Euclidean distance formula, which is:

[0079]

[0080] ④. Using the angle between the line connecting the UAV and the region vertex calculated in ② and the due north direction, calculate the angle between the line connecting the UAV and the region vertex and the horizontal field of view. The calculation formula is as follows:

[0081]

[0082] In the formula, hfov represents the horizontal field of view parameter of the camera on the UAV, yaw represents the yaw angle of the UAV, and angle represents the angle calculated by ②. angle This represents the angle between the line connecting the drone and the vertex of the region and the horizontal field of view.

[0083] ⑤. Using the distance calculated in ③, calculate the angle between the line connecting the UAV and the region vertex and the vertical field of view. The formula is:

[0084] h max =max(drone) height ,vertex height )

[0085] h min =min(drone) height ,vertex height )

[0086] delta h =h max -h min

[0087] theta = (180 / π) × arcsin(h / d)

[0088] In the formula, drone height ,vertex height Let a, b represent the altitudes of the drone and the region vertex, respectively. max(a, b) represents the maximum value of a and b, min(a, b) represents the minimum value of a and b, d is the straight-line distance between the drone and the region vertex calculated in step ②, arcsin represents the arcsine function, and delta... h The maximum height difference between the drone and the region's vertex is represented by the value of the 'theta', which indicates the angle between the line connecting the drone and the target and the vertical direction.

[0089] Then, based on the height relationship between the drone and the region vertex, the angle between the line connecting them and the vertical field of view is calculated. The calculation formula is as follows:

[0090]

[0091] In the formula, vfov represents the vertical field of view parameter of the camera on the UAV, pitch represents the pitch angle parameter of the UAV, theta represents the angle between the line connecting the UAV and the vertex of the region and the vertical direction, and y angle This represents the angle between the line connecting the drone and the region vertex and the vertical field of view.

[0092] ⑥. Calculate the pixel mapping values ​​on the x and y axes based on the included angles calculated in ④ and ⑤, respectively. The calculation formula is as follows:

[0093] x′=img weight *include angle / hfov

[0094] y′=img height *y angle / vfov

[0095] In the formula, img weight ,img height These represent the image width and height in the camera's resolution, respectively. angle y represents the angle between the line connecting the drone and the region vertex and the horizontal field of view. angle The angle between the line connecting the UAV and the region vertex and the vertical field of view is represented by hfov and vfov, which represent the horizontal and vertical field of view, respectively.

[0096] After calculating the mappings x′ and y′, determine whether their values ​​meet the requirements. The determination criteria are as follows:

[0097]

[0098] If the judgment passes, then the actual pixel value is mapped, and the calculation formula is as follows:

[0099] fx = img weight / 2×tan((π / 180)×hfov / 2)

[0100] fy = img height / 2×tan((π / 180)×vfov / 2)

[0101]

[0102] In the formula, tan represents the tangent function, fx represents the focal length on x, fy represents the focal length on y, and include angle The angle y represents the angle between the line connecting the drone and the region vertex and the horizontal field of view. angle This represents the angle between the line connecting the drone and the region vertex and the vertical field of view, shown in the image. weight ,img height represents the width and height of the camera resolution, respectively, and hfov and vfov represent the horizontal and vertical field of view parameters of the camera, respectively.

[0103] ⑦. Repeat steps ①-⑥ to obtain the pixel coordinates of all intrusion area vertices in the drone's view.

[0104] After obtaining the pixel coordinates of the intrusion area in the drone image, the target category is identified in real time from the video image using a target detection model, and the pixel coordinates of the target's detection box are output. The target detection model is a YOLO series algorithm, and the identified objects include, but are not limited to, people and vehicles. Specifically, in this embodiment, the YOLOv12-n model (lightweight deployment) is used to identify people / vehicle targets and output the coordinates of the detection box.

[0105] The Intersection over Union (IoU) ratio between the target detection bounding box and the intrusion region's pixel coordinates is calculated. If the IoU exceeds a preset threshold, it is determined to be an unauthorized intrusion. Specifically, the IoU (Intersection over Union) calculates the overlap ratio of the bounding rectangles of two objects, that is, the ratio of their intersection to their union. The calculation formula is as follows:

[0106]

[0107] The calculated IoU value is compared with a preset IoU threshold. If the IoU value is greater than the threshold, an unauthorized intrusion is detected, an alarm is issued to relevant personnel, and the information of the intruding target, time, and location are recorded to generate evidence data. If the IoU value is less than or equal to the threshold, monitoring continues. Alarm methods in this embodiment include audible and visual alarms, SMS notifications, and network message pushes. Using an IoU threshold mechanism can avoid false alarms due to minor overlap (such as personnel approaching the boundary), reducing the false positive rate.

[0108] Example 2

[0109] Example 2 is attached. Figure 3 As shown, a regional intrusion identification system from the perspective of an unmanned aerial vehicle (UAV) includes:

[0110] GIS area labeling module: Provides a map interactive interface for labeling intrusion areas.

[0111] UAV data acquisition module: Real-time acquisition of video streams and pose parameters.

[0112] Coordinate mapping module: Performs the conversion from geographic coordinates to image pixels.

[0113] Target recognition module: Detects targets based on the YOLO algorithm.

[0114] Intrusion detection module: Determines intrusion behavior through IoU calculation.

[0115] Alarm module: Triggers alarms in real time and stores evidence.

[0116] The system also applies one of the aforementioned methods for identifying regional intrusions from the perspective of a drone.

[0117] The above are merely embodiments of the present invention. Commonly known structures and characteristics are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A method for identifying regional intrusion from the perspective of an unmanned aerial vehicle (UAV), characterized in that, Including the following steps: S1: Mark the polygonal invasion area on the GIS map and obtain the geographic coordinate information of the area's vertices; S2: The drone collects video images and its own pose parameters in real time; S3: Based on the geographic coordinate information and pose parameters, calculate the pixel coordinates of the intrusion area in the UAV image using a spatial coordinate transformation algorithm; S4: Identify target categories from video images in real time using a target detection model, and output the pixel coordinates of the target's detection box; S5: Calculate the intersection-union ratio (IoU) between the target detection box and the pixel coordinates of the intrusion area. If the IoU exceeds the preset threshold, it is determined to be an unauthorized intrusion. S6: When unauthorized intrusion occurs, an alarm is issued to relevant personnel and evidence data is generated based on the video images used to determine the unauthorized intrusion.

2. The method for regional intrusion identification from the perspective of an unmanned aerial vehicle (UAV) according to claim 1, characterized in that, S1 includes: S11: Draw a closed polygonal intrusion area on a GIS map through interactive operations; S12: Stores geographic coordinate information for each vertex in the storage area, including latitude, longitude, and altitude.

3. The method for regional intrusion identification from the perspective of an unmanned aerial vehicle (UAV) according to claim 1, characterized in that, The spatial coordinate transformation algorithm in S3 includes: S31: Convert geographic coordinates to a 3D Cartesian coordinate system; S32: Combine the UAV pose parameters and the field of view of the camera on the UAV to calculate the pixel mapping value of the intrusion area vertex in the image; S33: Filter valid mapping coordinates by preset judgment conditions.

4. The method for regional intrusion identification from the perspective of an unmanned aerial vehicle (UAV) according to claim 3, characterized in that, The calculation of the pixel mapping value includes: S321: Calculate the angle between the target line and the horizontal / vertical field of view based on the straight-line distance, height difference, and camera parameters between the UAV and the region vertex; S322: Generate pixel coordinates based on the included angle and image resolution.

5. A method for regional intrusion identification from the perspective of an unmanned aerial vehicle (UAV) according to claim 4, characterized in that, S5 includes: S51: Calculate the overlap ratio (IoU) between the target detection bounding box and the intrusion region. S52: When the IoU exceeds the preset threshold, an unauthorized intrusion is determined.

6. A method for regional intrusion identification from the perspective of an unmanned aerial vehicle (UAV) according to claim 5, characterized in that: The target detection model is a YOLO series algorithm, and the objects to be identified include, but are not limited to, people and vehicles.

7. A method for regional intrusion identification from the perspective of an unmanned aerial vehicle (UAV) according to claim 6, characterized in that: The alarm methods of S6 include audible and visual alarms, SMS notifications, and network message pushes.

8. A regional intrusion identification system from the perspective of an unmanned aerial vehicle (UAV), characterized in that, include: GIS area labeling module: Provides a map interactive interface for labeling intrusion areas; UAV data acquisition module: acquires video streams and pose parameters in real time; Coordinate mapping module: Performs the conversion from geographic coordinates to image pixels; Target recognition module: Detects targets based on the YOLO algorithm; Intrusion detection module: Determines intrusion behavior through IoU calculation; Alarm module: Triggers alarms in real time and stores evidence.

9. A region intrusion identification system from the perspective of an unmanned aerial vehicle (UAV) according to claim 8, characterized in that: The method for identifying regional intrusion from the perspective of any one of the claims 1-7 is applied.

Images